White House keeps AI safety testing framework secret

The Trump administration finalized a framework for testing new AI models against security and cybersecurity risks following meetings with representatives from OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft. However, the White House decided not to publicly disclose the policy details, sharing testing criteria only with a select group of tech companies. The lack of transparency leaves open questions about the level of scrutiny models will face and the safety benchmarks they must meet. The framework was developed following the situation with Anthropic's Mythos model, which the company withheld from public release in April due to concerns it could be used to hack IT and financial systems.